Testing Mercury 3: Diffusion LLM Hits 1,068 tokens/sec, Mercury Decide Adds Vision
Sam Witteveen · youtube · 2026-10-20
Sam Witteveen tests Inception's diffusion LLM Mercury 3 (1,068 tokens/sec) and Mercury Decide, which can now read images.
Key points:
- How diffusion LLMs work; benchmarks vs GPT-6 Luna and Claude Haiku 5.5
- Swarms of sub-agents, 1M context, OpenAI-compatible API
- OcTen: search on smart glasses
- Real-time speed tests, 5,000-word essay, agentic tool-calling and RAG tests
- Mercury Decide image demo and System 1 / System 2 pattern
- Pipeline demo combining Mercury Decide + Mercury 3
- Launch pricing; free trial via OpenRouter
More from Models
- PretrainGym pits frontier agents on pretraining data curation: Kimi K3 wins at 1.042 — xeophon · 2026-10-20
- SlideOn model card details SAE pretraining and Matryoshka sparse training recipe — tomaarsen · 2026-10-20
- LightOn ships 16.3M training pairs and full code alongside SlideOn models — tomaarsen · 2026-10-20
- SlideOn lands in sentence-transformers with one-line SparseEncoder loading — tomaarsen · 2026-10-20
- mSlideOn handles cross-lingual retrieval with raw dot-product scoring — tomaarsen · 2026-10-20
- mSlideOn tops MTEB Code at 74.78 nDCG, beating dense and late-interaction models — tomaarsen · 2026-10-20